Robust Semantic Segmentation with Ladder-DenseNet Models
This work addresses semantic segmentation for computer vision applications, but it appears incremental as it applies a modified existing method to multiple datasets without claiming major breakthroughs.
The paper tackles semantic segmentation by evaluating a ladder-style DenseNet model on four benchmark datasets, achieving results that reveal interesting findings but without reporting specific performance numbers.
We present semantic segmentation experiments with a model capable to perform predictions on four benchmark datasets: Cityscapes, ScanNet, WildDash and KITTI. We employ a ladder-style convolutional architecture featuring a modified DenseNet-169 model in the downsampling datapath, and only one convolution in each stage of the upsampling datapath. Due to limited computing resources, we perform the training only on Cityscapes Fine train+val, ScanNet train, WildDash val and KITTI train. We evaluate the trained model on the test subsets of the four benchmarks in concordance with the guidelines of the Robust Vision Challenge ROB 2018. The performed experiments reveal several interesting findings which we describe and discuss.